4.7M trajectories on @base is the headline. But honestly, that’s not the part I find most interesting about @axisrobotics. It’s what happens when that much data starts compounding. More trajectories can improve the dataset. Better datasets can improve policies. Better policies create better real world signals. And those signals can shape what gets collected next. That creates something bigger than a growing dataset. It creates a feedback loop where yesterday’s contributions can make tomorrow’s data more valuable. That’s the part I’m watching going into September. Not how big the number gets. How much more useful the next million becomes than the last one.